List your company

Question Answering on TruthfulQA (MC1, Δ%)

27.54MC1 AccuracyDOLA dynamic
18.627220.941123.25525.5689Jun 6, 2026

Evaluation results

MethodLinks
DOLA dynamicModel=Qwen2.5-7B2026.0627.5426.4
CRYSTALBOOST (best)Model=Qwen2.5-7B2026.0622.282.2
BaselineModel=Qwen2.5-7B2026.0621.79—
CRYSTALBOOST (best)Model=Llama-3.1-8B2026.0620.568.4
DOLA dynamicModel=Llama-3.1-8B2026.0619.341.9
BaselineModel=Llama-3.1-8B2026.0618.97—